Adaptive filtering method, system, in-situ measurement device for semiconductor furnace tube temperature data

Abstract

The application provides a kind of semiconductor furnace tube temperature data self-adapting filtering method, system, in-situ measuring device, it is related to semiconductor detection technical field, the preset noise threshold value based on current furnace tube process window is introduced in the application, the fluctuation degree of temperature measurement result allowed by engineer can be quantified as the upper limit of filtering adjustment basis, so that filtering control process is simultaneously subjected to the common constraint of current signal quality and process requirement.Based on this, when the temperature measurement signal quality is poor, the temperature output fluctuation can be reduced by enhancing the filtering effect to improve the measurement accuracy and stability;When the temperature measurement signal quality is good or the process requires high dynamic tracking, the output lag can be reduced by weakening the filtering effect to improve the dynamic response performance, so as to realize the simultaneous optimization of dynamic response performance while ensuring the measurement accuracy and stability, and improve the adaptability of semiconductor furnace tube temperature measurement under different process windows.

Description

Technical Field

[0001] This application relates to the field of semiconductor testing technology, and in particular to an adaptive filtering method, system, and in-situ measurement device for semiconductor furnace tube temperature data. Background Technology

[0002] In semiconductor manufacturing processes, furnace tube equipment is crucial for key thermal processing steps such as diffusion, oxidation, low-pressure chemical vapor deposition (LPCVD), and atomic layer deposition (ALD). It typically includes diffusion furnaces, oxidation furnaces, LPCVD furnaces, and ALD furnaces. During these processes, the temperature inside the furnace tube must be precisely controlled within the range of 600°C to 1200°C. The accuracy of temperature control and the stability of measurement directly determine the uniformity of film thickness, doping concentration consistency, and device yield on the wafer.

[0003] In traditional technologies, furnace tube temperature measurement typically employs thermocouples or monochromatic infrared thermometry. Thermocouples, being contact-type temperature sensing elements, while relatively mature in structure, pose a risk of metal contamination in semiconductor manufacturing scenarios. Furthermore, they struggle to accurately reflect the true temperature of the wafer in applications requiring wafer rotation or non-contact measurement. While monochromatic infrared thermometry enables non-contact measurement, its results are susceptible to factors such as the absorption characteristics of process gases within the furnace, variations in radiation transmission under low-pressure conditions, and transmittance attenuation due to deposits on the quartz window surface. These factors lead to increased measurement errors, making it difficult to meet the precision and stability requirements of high-end semiconductor processes.

[0004] To address these issues, dual-color thermometers acquire thermal radiation signals at two wavelengths and calculate their emissivity ratio. Theoretically, this can reduce the impact of changes in optical transmittance and target emissivity fluctuations on the measurement results, and thus it has been used in the exploration of semiconductor furnace tube temperature measurement. However, the actual operating conditions of semiconductor furnace tubes differ significantly from general industrial dual-color temperature measurement applications: First, modern furnace tube processes typically have high heating and cooling rates, requiring the temperature measurement system to respond quickly to temperature changes to avoid overshoot or under-adjustment during temperature control. Second, the furnace tube contains multiple interference factors, including background radiation from the heating element, process gas disturbances, noise enhancement under low pressure, and signal attenuation due to window deposition, resulting in a low signal-to-noise ratio of the original dual-color temperature measurement signal, thus affecting the stability of the temperature calculation results. Third, semiconductor processes typically impose strict requirements on temperature measurement errors and repeatability; the output temperature cannot be severely delayed due to enhanced filtering, nor can noise fluctuations be amplified in pursuit of dynamic response.

[0005] Therefore, there is an urgent need for an adaptive filtering method suitable for semiconductor furnace tube temperature measurement data, which can dynamically adjust the length of the filtering window by combining the current signal quality, the process requirements of the current furnace tube process window, and process-related parameters, so as to optimize the dynamic response performance while ensuring measurement accuracy and stability. Summary of the Invention

[0006] The purpose of this application is to provide an adaptive filtering method, system, and in-situ measurement device for semiconductor furnace tube temperature data, so as to overcome the shortcomings of traditional technologies in that it is difficult to coordinate and optimize measurement accuracy, stability, and dynamic response.

[0007] In a first aspect, this application proposes an adaptive filtering method for semiconductor furnace tube temperature data, the method comprising: The system acquires a digital voltage signal, a preset noise threshold, and current process parameters. The digital voltage signal is obtained by photoelectric conversion and synchronous sampling of two reflected lights formed by two preset wavelength light signals irradiating the wafer surface inside the furnace tube. The preset noise threshold is set based on the process requirements of the current furnace tube process window. The current process parameters include process stage coefficients, quartz window cumulative deposition factor, and scaling factor. Calculate the instantaneous temperature sequence based on the digital voltage signal, and calculate the real-time signal-to-noise ratio based on the instantaneous temperature sequence; Calculate the target signal-to-noise ratio based on the preset noise threshold; The current filtering window length is determined based on the real-time signal-to-noise ratio, the target signal-to-noise ratio, and the current process parameters. The current filtering window length is associated with the ratio of the target signal-to-noise ratio to the real-time signal-to-noise ratio, and the current filtering window length is constrained and adjusted based on a preset window length threshold range. The instantaneous temperature sequence is filtered based on the adjusted current filter window length to obtain the final output temperature.

[0008] In one embodiment, calculating the instantaneous temperature sequence based on the digital voltage signal includes: Calculate the corresponding two-color ratio based on the two digital voltage signals at different sampling times; Based on the pre-stored calibration relationship between temperature and ratio, calculate the instantaneous temperature value corresponding to each of the two-color ratios; The instantaneous temperature values ​​are sorted according to the sampling time order to obtain the instantaneous temperature sequence.

[0009] In one embodiment, calculating the real-time signal-to-noise ratio based on the instantaneous temperature sequence includes: The instantaneous temperature sequence is processed in real time using the sliding window method to obtain the set of window temperatures corresponding to the current moment; The standard deviation of each instantaneous temperature value in the window temperature set relative to the window mean is calculated to obtain the temperature noise standard deviation; The real-time signal-to-noise ratio is calculated based on the preset scaling factor and the temperature noise standard deviation. The real-time signal-to-noise ratio is negatively correlated with the temperature noise standard deviation.

[0010] In one embodiment, calculating the target signal-to-noise ratio based on the preset noise threshold includes: The target standard deviation is determined based on the preset proportional relationship between the preset noise threshold and the noise standard deviation; The target signal-to-noise ratio is obtained by calculating the ratio of the preset scaling factor to the target standard deviation.

[0011] In one embodiment, the preset window length threshold includes a preset minimum window length and a preset maximum window length; The step of determining the current filter window length based on the real-time signal-to-noise ratio, the target signal-to-noise ratio, and the current process parameters includes: The candidate filter window length is determined by multiplying the process stage coefficient, the quartz window cumulative deposition factor, the proportional coefficient, and the ratio of the target signal-to-noise ratio to the real-time signal-to-noise ratio. Based on the preset minimum window length and the preset maximum window length, the candidate filter window length is subjected to amplitude limiting to obtain the current filter window length, which is expressed as:

[0012] in, The preset maximum window length; Set the minimum window length; This is the proportionality coefficient; This is the process stage coefficient; For the quartz window, the accumulation deposition factor; The target signal-to-noise ratio; For real-time signal-to-noise ratio; It is a non-zero positive number, used to avoid division by zero; This indicates rounding down to the nearest integer.

[0013] In one embodiment, the step of limiting the candidate filter window length based on a preset minimum window length and a preset maximum window length to obtain the current filter window length includes: If the candidate filter window length is less than the preset minimum window length, the preset minimum window length shall be used as the current filter window length. If the candidate filter window length is greater than the preset maximum window length, the preset maximum window length shall be used as the current filter window length. If the candidate filter window length is between the preset minimum window length and the preset maximum window length, the candidate filter window length is taken as the current filter window length.

[0014] In one embodiment, the method further includes: According to the preset monitoring cycle, acquire the data of the final output temperature within the most recent monitoring window, and calculate the output standard deviation based on the data; If the output standard deviation is greater than a first preset multiple of the target standard deviation, the proportional coefficient is increased according to a preset ratio. If the output standard deviation is less than a second preset multiple of the target standard deviation and the real-time signal-to-noise ratio is continuously lower than the target signal-to-noise ratio, the proportional coefficient is reduced according to a preset ratio. Otherwise, keep the aforementioned proportionality coefficient unchanged; Based on the adjusted scaling factor, the filter window length for subsequent time steps is redetermined.

[0015] In one embodiment, filtering the instantaneous temperature sequence according to the current filter window length to obtain the final output temperature includes: Based on the current filter window length, the instantaneous temperature sequence is subjected to a moving average filter to obtain the candidate output temperature at the current moment; If the current filter window length is inconsistent with the filter window length of the previous time step, a weighted update is performed based on the output temperature of the previous time step and the candidate output temperature to obtain the final output temperature of the current time step. If the current filter window length is the same as the filter window length at the previous moment, the candidate output temperature is taken as the final output temperature at the current moment.

[0016] In one embodiment, the method further includes: If the real-time signal-to-noise ratio remains below the preset lower limit for a preset duration, it is determined that the current signal quality does not meet the measurement requirements, and an alarm signal is output.

[0017] Secondly, this application proposes an adaptive filtering system for semiconductor furnace tube temperature data, the system comprising: The acquisition module is used to acquire digital voltage signals of two preset wavelength band measurement optical signals, a preset noise threshold, and current process parameters; wherein, the digital voltage signals are obtained by synchronously sampling two reflected lights on the wafer surface inside the furnace tube, and the reflected lights are generated based on the optical signals of two preset wavelength bands illuminating the wafer surface; the preset noise threshold is set based on the process requirements of the current furnace tube process window; The calculation module is used to calculate the instantaneous temperature sequence based on the digital voltage signal, and calculate the real-time signal-to-noise ratio based on the instantaneous temperature sequence; calculate the target signal-to-noise ratio based on the preset noise threshold; and determine the current filter window length based on the real-time signal-to-noise ratio, the target signal-to-noise ratio, and the current process parameters. The filtering module is used to filter the instantaneous temperature sequence according to the current filtering window length to obtain the final output temperature.

[0018] Thirdly, this application proposes an in-situ measurement device, the device comprising: The light source module is used to generate measurement light, separate the measurement light into two preset wavelength bands, and irradiate the wafer surface inside the furnace tube. The acquisition module is used to simultaneously sample two reflected lights generated on the wafer surface based on the optical signal and convert them into digital voltage signals; The adaptive filtering system for semiconductor furnace tube temperature data as described in the second aspect is used to calculate an instantaneous temperature sequence based on the digital voltage signal and to calculate a real-time signal-to-noise ratio based on the instantaneous temperature sequence; to calculate a target signal-to-noise ratio based on the preset noise threshold; to determine a current filtering window length based on the real-time signal-to-noise ratio, the target signal-to-noise ratio, and the current process parameters; and to filter the instantaneous temperature sequence based on the current filtering window length to obtain the final output temperature.

[0019] Fourthly, this application also provides a computer device. The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the method steps of the first aspect.

[0020] Fifthly, this application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, which, when executed by a processor, implements the method steps of the first aspect.

[0021] The aforementioned adaptive filtering method, system, and in-situ measurement device for semiconductor furnace tube temperature data have at least the following advantages: This application calculates the instantaneous temperature sequence and real-time signal-to-noise ratio (SNR) based on the digital voltage signal, calculates the target SNR based on a preset noise threshold, and then dynamically determines the current filtering window length based on the real-time SNR, target SNR, and current process parameters. The instantaneous temperature sequence is then filtered based on the adjusted filtering window length. This application introduces a preset noise threshold set based on the process requirements of the current furnace tube process window. This quantifies the upper limit of the allowable deviation of the temperature measurement result fluctuation for engineers as the basis for filtering adjustment, making the filtering control process simultaneously constrained by both the current signal quality and process requirements. Based on this, when the temperature measurement signal quality is poor, the temperature output fluctuation can be reduced by enhancing the filtering effect to improve measurement accuracy and stability. When the temperature measurement signal quality is good or the process has high requirements for dynamic tracking, the output hysteresis can be reduced by weakening the filtering effect to improve dynamic response performance. This achieves synergistic optimization of dynamic response performance while ensuring measurement accuracy and stability, and improves the adaptability of semiconductor furnace tube temperature measurement under different process windows. Attached Figure Description

[0022] Figure 1 This is a structural block diagram of an in-situ measurement device in one embodiment; Figure 2 This is a flowchart illustrating an adaptive filtering method for semiconductor furnace tube temperature data in one embodiment. Figure 3 This is a flowchart illustrating the steps for calculating the real-time signal-to-noise ratio in one embodiment; Figure 4 This is a flowchart illustrating the steps for calculating the target signal-to-noise ratio in one embodiment; Figure 5 This is a block diagram of an adaptive filtering system for semiconductor furnace tube temperature data in one embodiment. Detailed Implementation

[0023] The following specific examples illustrate the implementation of this application. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. This application can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this application. It should be noted that, unless otherwise specified, the following embodiments and features in the embodiments can be combined with each other.

[0024] Some exemplary embodiments of this application have been described for illustrative purposes. It should be understood that this application may be implemented in other ways not specifically shown in the accompanying drawings.

[0025] Please see Figure 1In one exemplary embodiment, this application provides an in-situ measurement device, including: a light source module, a data acquisition module, and an adaptive filtering system for semiconductor furnace tube temperature data.

[0026] The light source module is used to generate measurement light, which is then separated into two preset wavelength signals and irradiated onto the wafer surface inside the furnace tube.

[0027] Optionally, the light source module in this embodiment includes a broadband light source, a front protection window, a front lens, a beam splitter, a first narrowband filter, and a second narrowband filter.

[0028] Specifically, a broadband light source is used to generate measurement light covering a first preset wavelength band and a second preset wavelength band, which enters the furnace tube through a front protection window; a front lens is used to collimate and focus the measurement light; a beam splitter is used to separate the measurement light into two optical signals. Exemplarily, the beam splitter can be a dichroic mirror, a beam splitter prism, or a grating. A first narrowband filter and a second narrowband filter are respectively disposed on corresponding optical paths to select the first preset wavelength band and the second preset wavelength band from the two optical signals, respectively. Exemplarily, the center wavelengths of the first and second preset wavelength bands can be 0.9 μm and 1.05 μm, respectively, and the bandwidths can both be 50 nm. The separated optical signals of the two preset wavelength bands are then irradiated onto the wafer surface inside the furnace tube through a beam combining optical path and an exiting optics, or irradiated onto the same measurement area on the wafer surface along a near-coaxial optical path, so as to subsequently acquire the corresponding reflected light signals.

[0029] The acquisition module is used to simultaneously sample two reflected lights generated by optical signals on the wafer surface and convert them into digital voltage signals.

[0030] Optionally, the acquisition module includes two photodetectors and a signal conditioning unit, the signal conditioning unit including a preamplifier, a programmable gain amplifier (PGA), an anti-aliasing filter and an analog-to-digital converter (ADC) connected in sequence.

[0031] Specifically, two photodetectors receive two reflected lights from the wafer surface and convert the corresponding optical signals into electrical signals. The photodetectors can be InGaAs or Si detectors; in other embodiments, a TEC (Thermal Control Unit) temperature control component can be configured to stabilize the detector's operating temperature, reduce dark current drift and device response fluctuations, and improve measurement stability.

[0032] The signal conditioning unit is electrically connected to two photodetectors and is used to condition and acquire the weak electrical signals output by the detectors. During operation, the preamplifier amplifies the weak current or voltage signals output by the detectors; the programmable gain amplifier adaptively adjusts the amplification factor according to the signal strength to improve the dynamic range utilization under different operating conditions; the anti-aliasing filter suppresses high-frequency noise and out-of-band interference; and the analog-to-digital converter synchronously samples and converts the conditioned analog signal to digital voltage signals corresponding to two preset frequency bands. and For example, the analog-to-digital converter has a resolution of at least 16 bits and an adjustable sampling rate, typically 100Hz.

[0033] In one example, the sampling process of the two channels is controlled by the same clock source to ensure that the reflected signals of the two bands are acquired at the same sampling time, thereby improving the accuracy of subsequent data processing and instantaneous temperature calculation.

[0034] An adaptive filtering system for semiconductor furnace tube temperature data is used to calculate the instantaneous temperature sequence based on the digital voltage signal, and to calculate the real-time signal-to-noise ratio (SNR) based on the instantaneous temperature sequence; to calculate the target SNR based on a preset noise threshold; to determine the current filtering window length based on the real-time SNR, the target SNR, and the current process parameters; and to filter the instantaneous temperature sequence based on the current filtering window length to obtain the final output temperature. The preset noise threshold is set based on the process requirements of the current furnace tube process window.

[0035] Optionally, the above-mentioned in-situ measurement device further includes a human-machine interaction unit.

[0036] The human-machine interface unit is used to display temperature data after adaptive filtering and to receive process setting parameters and preset noise thresholds input by engineers.

[0037] For example, the engineer can input a preset noise threshold corresponding to the current furnace tube process window through the human-computer interaction unit. In this embodiment, the preset noise threshold can be the maximum allowable peak-to-peak output noise level. The value is expressed in °C. For example, in a polysilicon deposition process, the maximum allowable peak-to-peak output noise level can be set to 1.0 °C; in an oxide deposition process, the maximum allowable peak-to-peak output noise level can be set to 1.5 °C. After receiving this parameter, the system writes it to non-volatile memory for storage, so that the corresponding settings are retained even after the device is powered off.

[0038] Furthermore, the human-machine interface unit can also support process recipe management functions. Different process recipes can correspond to different preset noise thresholds, filtering parameters, and alarm thresholds. When an engineer selects the corresponding process recipe, the system can automatically call the preset noise thresholds associated with that process recipe and perform subsequent target signal-to-noise ratio calculations and filter window length adjustments accordingly, thereby improving the adaptability of temperature measurement and filtering control under different furnace tube process conditions.

[0039] The aforementioned in-situ measurement device provides two preset wavelengths of measurement light to the wafer surface inside the furnace tube via a light source module. A data acquisition module simultaneously acquires the corresponding reflected light to form digital voltage signals. These signals are then combined with an adaptive filtering system for the semiconductor furnace tube temperature data. Based on a preset noise threshold, real-time signal-to-noise ratio, and current process parameters, adaptive filtering is performed, enabling in-situ measurement of wafer temperature under complex conditions within the semiconductor furnace tube. This application avoids the risks of metal contamination and contact scratches associated with traditional contact temperature measurement methods, as well as the effects of emissivity fluctuations, window transmittance changes, and process atmosphere disturbances in conventional monochromatic infrared temperature measurement methods. Furthermore, this application dynamically adjusts the filtering intensity according to different process stages and signal quality, simultaneously optimizing the dynamic response performance to temperature changes while ensuring measurement accuracy and stability, thus improving the reliability of furnace tube temperature monitoring. Please see Figure 2 In one exemplary embodiment, this application provides an adaptive filtering method for semiconductor furnace tube temperature data, specifically including the following steps: Step 202: Acquire digital voltage signal, preset noise threshold and current process parameters; wherein, the digital voltage signal is obtained by photoelectric conversion and synchronous sampling of two reflected lights formed by two preset wavelength light signals irradiating the wafer surface inside the furnace tube; the preset noise threshold is set based on the process requirements of the current furnace tube process window; the current process parameters include process stage coefficient, quartz window cumulative deposition factor and scaling factor.

[0040] Specifically, the acquisition module in the in-situ measurement device uses a fixed sampling rate. The reflected light signals of the two bands are sampled synchronously in real time to obtain the digital voltage signals of the two bands. and Among them, sampling rate The frequency can be set according to the rate of change of furnace tube temperature, for example, 20Hz to 50Hz, to optimize the temperature change tracking capability and signal acquisition stability.

[0041] The digital voltage signal is a discrete-time signal obtained by photoelectric conversion, amplification, conditioning, and analog-to-digital conversion of two reflected lights formed by two preset wavelength signals irradiating the wafer surface inside the furnace tube. Specifically, the first wavelength reflected light is converted into a first digital voltage signal after passing through the first photoelectric detection channel. The reflected light from the second band is converted into a second digital voltage signal after passing through the second photoelectric detection channel. The two channels are preferably controlled by the same clock source for synchronous sampling to ensure the timing consistency of subsequent two-color ratio calculation and temperature inversion.

[0042] The preset noise threshold is used to characterize the range of output temperature fluctuations allowed under the current furnace tube process window. It can be preset by engineers according to specific process requirements and stored in non-volatile memory after being input through the human-machine interaction unit.

[0043] The current process parameters reflect the impact of the current furnace tube operating conditions on the determination of the filter window length, and may include process stage coefficients. Quartz window cumulative deposition factor and proportionality coefficient .

[0044] Among them, the process stage coefficient This is used to characterize the different requirements of temperature response speed and filtering intensity at different process stages. By reading the current process stage flag from the furnace tube main controller in real time, the process stage coefficient at the current moment can be determined according to preset rules. For example, it can be determined according to the following rule: if it is in the heating phase, then it is considered... The value ranges from 0.5 to 0.8; if it is in the isothermal deposition stage, then If it is in the cooling phase, then The value ranges from 0.8 to 1.2.

[0045] The above scheme is adopted because the heating stage needs to track temperature changes more quickly, and the window length can be suppressed by a smaller stage coefficient; the isothermal deposition stage has higher stability requirements, so the stage coefficient can be set to a baseline value of 1; the cooling stage can be appropriately adjusted according to actual process requirements to synergistically optimize the dynamic response and noise resistance during the cooling process.

[0046] Furthermore, the quartz window accumulation deposition factor This is used to characterize the degree of optical signal attenuation caused by deposits accumulating on the surface of a quartz window. It can be determined using either of the following two methods: The first method is an empirical model estimation based on cumulative process runtime. Specifically, it records the cumulative process runtime since the quartz window was installed or the last cleaning was completed. And according to the preset deposition rate coefficient calculate:

[0047] in, It can be preset based on historical process data, equipment calibration results, or empirical statistics. As the cumulative process running time increases, The value gradually increases to reflect the signal quality degradation caused by the decrease in window transmittance.

[0048] The second method is a real-time estimation based on the attenuation ratio of the original signal strength. Specifically, the average intensity of the original digital voltage signals in two bands within a preset time window is monitored and compared with the initial reference intensity recorded under clean conditions to obtain the current signal strength attenuation ratio. The quartz window cumulative deposition factor can be determined as:

[0049] in, This represents the baseline average signal strength under clean window conditions. This represents the average signal strength at the current time or within the current statistical window. To prevent parameter loss of control under abnormal operating conditions, it can also be... Set an upper limit, for example, to The maximum value is limited to 3.

[0050] proportionality coefficient A scaling factor used to adjust the mapping between the target signal-to-noise ratio, the real-time signal-to-noise ratio, and the filter window length. Values ​​can be preset or stored and retrieved separately according to different process formulas. In some implementations, the proportional coefficient... Furthermore, online fine-tuning can be performed by combining the output noise closed-loop monitoring results to further improve the filter control accuracy. For example, The initial value can be set between 1.5 and 5.

[0051] Step 204: Calculate the instantaneous temperature sequence based on the digital voltage signal, and calculate the real-time signal-to-noise ratio based on the instantaneous temperature sequence.

[0052] Specifically, an instantaneous temperature sequence refers to a temperature data sequence formed by arranging the instantaneous temperature values ​​calculated at each sampling time based on digital voltage signals in two bands in chronological order. This instantaneous temperature sequence is used to characterize the raw measurement results of the temperature change of the wafer surface inside the furnace tube over time.

[0053] Real-time signal-to-noise ratio (SNR) is a parameter used to characterize the quality of the current temperature measurement signal, determined based on the fluctuations of the instantaneous temperature sequence within the current time period. A higher SNR indicates less noise and better signal quality in the current temperature measurement result; a lower SNR indicates more significant fluctuations and greater noise impact in the current temperature measurement result. In this embodiment, the instantaneous temperature sequence serves as the input basis for subsequent filtering processing, and the real-time SNR is used as an evaluation index to measure the quality of the current measurement data, subsequently determining the filtering strength suitable for the current operating conditions. Step 206: Calculate the target signal-to-noise ratio based on the preset noise threshold.

[0054] Specifically, the preset noise threshold refers to the allowable output noise range pre-set based on the process requirements of the current furnace tube process window, used to characterize the upper limit of the allowable deviation of the temperature measurement result fluctuation for engineers.

[0055] The target signal-to-noise ratio (SNR) is a target signal quality index calculated based on the preset noise threshold, used to reflect the stability level that the temperature measurement signal should achieve under the current process requirements. In this embodiment, the target SNR is used as a reference benchmark for subsequent adaptive adjustment of filtering parameters, so that the filtered temperature output meets the noise control requirements corresponding to the current process window.

[0056] Step 208: Determine the current filtering window length based on the real-time signal-to-noise ratio, the target signal-to-noise ratio, and the current process parameters. Filter the instantaneous temperature sequence according to the adjusted current filtering window length to obtain the final output temperature. The current filtering window length is related to the ratio of the target signal-to-noise ratio to the real-time signal-to-noise ratio, and the current filtering window length is constrained and adjusted based on a preset window length threshold range.

[0057] Specifically, the current filter window length refers to the number of instantaneous temperature values ​​participating in the temperature filtering calculation at the current moment, used to characterize the current filtering intensity. Generally, the larger the current filter window length, the higher the filtering intensity and the more stable the output temperature; the smaller the current filter window length, the lower the filtering intensity and the faster the output temperature responds to temperature changes.

[0058] The ratio of the target signal-to-noise ratio (SNR) to the real-time SNR reflects the degree of deviation between the current temperature measurement signal quality and the target signal quality. Current process parameters reflect the impact of the current process stage, window deposition state, and proportional adjustment requirements on the filtering intensity. The current filtering window length is determined based on the real-time SNR, target SNR, and current process parameters, and is adjusted within a preset window length threshold range to avoid excessive output fluctuations due to an excessively small filtering window or excessive output lag due to an excessively large filtering window. Filtering the instantaneous temperature sequence based on the adjusted current filtering window length yields a temperature output result adapted to the current process state. The aforementioned adaptive filtering method for semiconductor furnace tube temperature data calculates the instantaneous temperature sequence and real-time signal-to-noise ratio (SNR) based on the digital voltage signal, calculates the target SNR based on a preset noise threshold, and then dynamically determines the current filtering window length based on the real-time SNR, target SNR, and current process parameters. The instantaneous temperature sequence is then filtered based on the adjusted filtering window length. This application introduces a preset noise threshold set based on the process requirements of the current furnace tube process window. This quantifies the upper limit of the allowable deviation of the temperature measurement result fluctuation for engineers as the basis for filtering adjustment, ensuring that the filtering control process is simultaneously constrained by both the current signal quality and process requirements. Therefore, when the temperature measurement signal quality is poor, the filtering effect can be enhanced to reduce temperature output fluctuations, thereby improving measurement accuracy and stability. When the temperature measurement signal quality is good or the process has high requirements for dynamic tracking, the filtering effect can be weakened to reduce output hysteresis, thereby improving dynamic response performance. This achieves synergistic optimization of dynamic response performance while ensuring measurement accuracy and stability, and improves the adaptability of semiconductor furnace tube temperature measurement under different process windows.

[0059] Optionally, the instantaneous temperature sequence is calculated based on the digital voltage signal, including: Based on the two digital voltage signals at different sampling times, calculate the corresponding two-color ratio; based on the pre-stored calibration relationship between temperature and ratio, calculate the instantaneous temperature value corresponding to each two-color ratio; sort the instantaneous temperature values ​​according to the sampling time order to obtain the instantaneous temperature sequence.

[0060] Specifically, the basic principle of a dual-color thermometer is that the thermal radiation intensity of the same temperature measurement target at two different wavelengths is related to the target temperature. By obtaining the intensity ratio of the radiation signals in the two bands, the influence of changes in the overall transmittance of the optical path, slow fluctuations in the target emissivity, and changes in the detection gain on the temperature measurement results can be reduced, thereby improving the stability of temperature measurement.

[0061] For the wafer surface inside the furnace tube in this embodiment, the acquisition module simultaneously acquires digital voltage signals corresponding to two preset wavelength bands. and ,in, Indicates time The digital voltage signal in the first band, Indicates time The digital voltage signal in the second band.

[0062] For example, the corresponding two-color ratio value can be calculated based on two digital voltage signals at different sampling times, and can be expressed as:

[0063] in, Let be the two-color ratio at time t.

[0064] Furthermore, after obtaining the two-color ratio Then, the pre-stored temperature-ratio calibration relationship is invoked to calculate the corresponding instantaneous temperature value. The temperature-ratio calibration relationship can be pre-established through a standard blackbody calibration experiment and stored in the form of a lookup table, piecewise function, or fitted function. For example, the instantaneous temperature value can be expressed as:

[0065] in, This represents the temperature-ratio mapping function.

[0066] Based on the above process, for different sampling times The two digital voltage signals acquired synchronously can be used to calculate the corresponding two-color ratio. And further obtain the corresponding instantaneous temperature value. .

[0067] Subsequently, the instantaneous temperature values ​​are arranged according to the sampling time sequence to form an instantaneous temperature sequence. .

[0068] Please see Figure 3 Optionally, the real-time signal-to-noise ratio is calculated based on the instantaneous temperature sequence, including: Step 302: The instantaneous temperature sequence is processed in real time using the sliding window method to obtain the set of window temperatures corresponding to the current moment.

[0069] Step 304: Calculate the standard deviation of each instantaneous temperature value in the window temperature set relative to the window mean to obtain the temperature noise standard deviation.

[0070] Step 306: Calculate the real-time signal-to-noise ratio at the current moment based on the preset scaling factor and the temperature noise standard deviation. The real-time signal-to-noise ratio is negatively correlated with the temperature noise standard deviation.

[0071] Specifically, the real-time signal-to-noise ratio is used to characterize the noise level and signal quality of the current temperature measurement data.

[0072] For example, the length is set to A sliding buffer used to store the most recent instantaneous temperature value .in, It can be set according to the sampling rate and process response requirements, for example, it can be taken as follows: .

[0073] Whenever a new instantaneous temperature value is acquired, it is written to the sliding buffer, and the earliest instantaneous temperature value to enter the buffer is removed, thus achieving the tracking of the most recent... The real-time update of each instantaneous temperature value yields the set of window temperatures corresponding to the current moment, expressed as: .

[0074] The aforementioned window temperature set includes the current time and the consecutive W values ​​prior to it. The instantaneous temperature value corresponding to one sampling moment.

[0075] Furthermore, calculate the window mean of the instantaneous temperature values ​​within the current window. Its expression is:

[0076] Then, calculate the standard deviation of the instantaneous temperature value within the current window relative to the window mean. Its expression is:

[0077] in, Indicates the current time The temperature noise standard deviation is used to reflect the degree of fluctuation of the instantaneous temperature sequence within the current window.

[0078] Furthermore, based on the preset scaling factor and the temperature noise standard deviation, the real-time signal-to-noise ratio at the current moment is calculated, and its expression is:

[0079] in, This is a preset scaling factor used to map the standard deviation of temperature noise to a uniform signal-to-noise ratio evaluation scale.

[0080] From the above formula, we can see that the real-time signal-to-noise ratio With temperature noise standard deviation Negative correlation, meaning that the smaller the standard deviation of temperature noise, the higher the real-time signal-to-noise ratio; the larger the standard deviation of temperature noise, the lower the real-time signal-to-noise ratio.

[0081] The above scheme uses a sliding window method to perform online statistical analysis of the instantaneous temperature sequence and utilizes the standard deviation of temperature fluctuations within the window to characterize the current noise level in real time, which can reflect the changes in the quality of the current temperature measurement data in a timely manner. At the same time, by further converting the temperature noise standard deviation into a real-time signal-to-noise ratio, an intuitive evaluation basis can be provided for the adaptive adjustment of the subsequent filter window length. This allows the filter intensity to be dynamically adjusted according to changes in the current signal quality, suppressing temperature output fluctuations and improving measurement stability while avoiding response lag caused by over-filtering. This is beneficial for the synergistic optimization of measurement accuracy, stability, and dynamic response performance.

[0082] Please see Figure 4Optionally, the target signal-to-noise ratio is calculated based on a preset noise threshold, including: Step 402: Determine the target standard deviation based on the preset ratio between the preset noise threshold and the noise standard deviation.

[0083] Step 404: Calculate the ratio of the preset scaling factor to the target standard deviation to obtain the target signal-to-noise ratio.

[0084] Specifically, after obtaining the preset noise threshold, this application further calculates the target signal-to-noise ratio based on the preset noise threshold, so as to convert the upper limit of the allowable deviation of the temperature output fluctuation by the engineer into the target signal quality index on which the subsequent filtering control is based.

[0085] In this embodiment, the preset noise threshold is the maximum allowable peak-to-peak output noise level. The unit is °C. Since temperature output noise can be approximated as Gaussian white noise under certain conditions, based on the statistical relationship between peak-to-peak value and standard deviation: The preset noise threshold can be converted into the target standard deviation. .in, This represents the peak-to-peak value of the noise level. This represents the standard deviation of noise.

[0086] Furthermore, based on the preset proportional relationship between the preset noise threshold and the noise standard deviation, the target standard deviation is determined, and its expression is:

[0087] in, The target standard deviation is used to characterize the level of temperature output fluctuation allowed under the current process window requirements.

[0088] Furthermore, the ratio of the preset scaling factor to the target standard deviation is calculated to obtain the target signal-to-noise ratio, which is expressed as:

[0089] in, Indicates the target signal-to-noise ratio. This represents the preset scaling factor. It should be noted that this preset scaling factor is preferably consistent with the scaling factor used in the real-time signal-to-noise ratio (SNR) calculation to ensure that the target SNR and the real-time SNR are evaluated on a unified scale. For example, The possible value is 20.

[0090] Therefore, the target signal-to-noise ratio and the preset noise threshold are related. The correlation is negative; that is, the smaller the maximum allowable peak-to-peak value of the output noise set by the engineer, the higher the target signal-to-noise ratio (SNR); conversely, the larger the maximum allowable peak-to-peak value of the output noise set by the engineer, the lower the target SNR. This allows the different requirements for the degree of fluctuation in temperature measurement results under different process windows to be uniformly converted into the target signal quality index used for subsequent filter parameter adjustment.

[0091] By adopting the above scheme, the preset noise threshold set by the engineer is converted into the target standard deviation and then further converted into the target signal-to-noise ratio. This allows the upper limit of the allowable deviation of the temperature measurement result fluctuation to be quantified into the filter control target, thus providing a unified basis for subsequent adjustment of the filter window length. At the same time, since the target signal-to-noise ratio and the real-time signal-to-noise ratio use the same evaluation scale, it is easy to compare the two directly. This facilitates the dynamic adjustment of the filter intensity based on the current signal quality and process requirements, thereby ensuring measurement accuracy and stability while synergistically optimizing dynamic response performance and improving the adaptability and reliability of temperature measurement under different process windows.

[0092] Optionally, when the preset window length threshold includes a preset minimum window length and a preset maximum window length, the current filtering window length is determined based on the real-time signal-to-noise ratio, the target signal-to-noise ratio, and the current process parameters, including: The candidate filtering window length is determined by multiplying the process stage coefficient, the quartz window cumulative deposition factor, the proportionality coefficient, and the ratio of the target signal-to-noise ratio to the real-time signal-to-noise ratio. Based on the preset minimum and maximum window lengths, the candidate filtering window length is limited to obtain the current filtering window length, expressed as:

[0093] in, The preset maximum window length; Set the minimum window length; This is the proportionality coefficient; This is the process stage coefficient; For the quartz window, the accumulation deposition factor; The target signal-to-noise ratio; For real-time signal-to-noise ratio; It is a non-zero positive number, used to avoid division by zero; This indicates rounding down to the nearest integer.

[0094] Specifically, as shown in the above formula, when the real-time signal-to-noise ratio At higher levels, The current filter window length is reduced accordingly. Tends to the preset minimum window length This reduces the filtering intensity and improves the output temperature response speed to temperature changes; when the real-time signal-to-noise ratio At lower levels, The current filter window length will be increased accordingly. The filter capacity is also increased accordingly, thereby enhancing the filtering effect and suppressing temperature output fluctuations.

[0095] At the same time, when a faster response is required at the process stage, a smaller process stage coefficient can be used. Suppressing the increase in window length; when quartz window deposition is severe, a larger quartz window can accumulate deposition factors. Increasing the window length compensates for the noise impact caused by signal attenuation. In other words, the real-time signal-to-noise ratio, process stage, and window deposition state jointly determine the final filtering strength.

[0096] Furthermore, in a feasible embodiment, if the real-time signal-to-noise ratio Much higher than the target signal-to-noise ratio And process stage coefficient With quartz window cumulative deposition factor If the product is close to 1, it indicates that the current signal quality is good, and the need for correction of filtering enhancement due to the process stage and window state is weak. In this case, the current filtering window length can be directly determined as the preset minimum window length. To achieve faster dynamic response.

[0097] Optionally, based on a preset minimum window length and a preset maximum window length, the candidate filter window length is limited to obtain the current filter window length, including: If the candidate filter window length is less than the preset minimum window length, the preset minimum window length is used as the current filter window length; if the candidate filter window length is greater than the preset maximum window length, the preset maximum window length is used as the current filter window length; if the candidate filter window length is between the preset minimum window length and the preset maximum window length, the candidate filter window length is used as the current filter window length.

[0098] By adopting the above scheme, a multi-dimensional mapping relationship is constructed between the real-time signal-to-noise ratio, the target signal-to-noise ratio, and the current process parameters to the filter window length. Combined with the minimum and maximum window lengths for amplitude limiting constraints, the determination of the current filter window length can take into account both the deviation between the current signal quality and the target signal quality, and the requirements of the process stage for dynamic response and the influence of the quartz window deposition state on signal attenuation, thereby making the filter strength more matched with the actual working conditions.

[0099] Furthermore, when the signal quality is good, the current filter window length can automatically tend towards a smaller value to reduce output hysteresis and improve dynamic response performance; when the signal quality is poor or window deposition is severe, the current filter window length can automatically increase to enhance the filtering effect, reduce temperature output fluctuations, and improve measurement accuracy and stability. Therefore, this application can more effectively synergistically optimize measurement accuracy, output stability, and dynamic response performance under complex semiconductor furnace tube operating conditions, improving the adaptability and reliability of temperature measurement and filtering control under different process windows.

[0100] Optionally, the adaptive filtering method for the semiconductor furnace tube temperature data mentioned above further includes: According to the preset monitoring cycle, the data of the final output temperature within the most recent monitoring window is acquired, and the output standard deviation is calculated based on the data. If the output standard deviation is greater than a first preset multiple of the target standard deviation, the scaling factor is increased according to a preset ratio. If the output standard deviation is less than a second preset multiple of the target standard deviation and the real-time signal-to-noise ratio is continuously lower than the target signal-to-noise ratio, the scaling factor is decreased according to a preset ratio. Otherwise, the scaling factor remains unchanged. Based on the adjusted scaling factor, the filter window length for subsequent time periods is re-determined.

[0101] Specifically, after obtaining the final output temperature, this embodiment can also introduce a closed-loop feedback fine-tuning mechanism to adjust the scaling factor. Online corrections are performed to further reduce the deviation between the theoretical mapping relationship and the actual furnace tube operating conditions, thereby improving the accuracy of filter control.

[0102] For example, this application follows a preset monitoring cycle. For the final output temperature Periodic monitoring will be conducted. The monitoring period will be preset. The timeout can be set according to the rate of change of quartz window deposition and process conditions, for example, 30 to 60 seconds, to adapt to the slow-changing characteristics of window deposition. At the end of each monitoring cycle, the final output temperature data within the most recent monitoring window is acquired, and the output standard deviation is calculated based on the data within that monitoring window. For example, at the current monitoring time If the number of sampling points corresponding to the monitoring window length is... Then the expression for the set of output temperatures within the monitoring window is:

[0103] Calculate the average value of the final output temperature within the monitoring window. Its expression is:

[0104] Calculate the standard deviation of the final output temperature relative to the mean within the monitoring window. Its expression is:

[0105] in, This standard deviation is used to characterize the actual fluctuation of the filtered output temperature within the current monitoring window. The larger the standard deviation, the more significant the fluctuation in the filtered temperature output; the smaller the standard deviation, the more stable the filtered temperature output.

[0106] After obtaining the output standard deviation Then, compare it with the target standard deviation. Compare the target standard deviation. It is calculated from a preset noise threshold and used as the target for allowable output fluctuation under the current process window.

[0107] In a feasible embodiment, the following closed-loop regulation rule can be adopted: When satisfied When the actual output noise is determined to be too high, enhanced filtering is required; among which, The first preset multiple is exemplarily acceptable. At this point, the proportional coefficient is increased according to the preset ratio. For example, increase by 1.1 times.

[0108] When satisfied And real-time signal-to-noise ratio consistently below the target signal-to-noise ratio At that time, it was determined that the current filtering effect was too strong, showing a tendency of over-filtering; among them, For example, a second preset multiple can be taken. It should be noted that "consistently lower than" here means that the real-time signal-to-noise ratio is lower than the target signal-to-noise ratio for a preset continuous judgment period, or for a number of recent sampling times. In this case, the scaling factor is reduced according to a preset ratio. For example, a reduction of 0.9 times.

[0109] In all other cases, maintain the proportionality coefficient. constant.

[0110] Optionally, to avoid excessive increase or decrease of the proportional coefficient during closed-loop adjustment, this application also adjusts the proportional coefficient... The range of values ​​can be restricted. For example, it can be restricted to:

[0111] in, and These are the lower and upper limits of the proportionality coefficient, respectively. For example, 1.5 is acceptable. A value of 5 is acceptable. This limitation prevents excessive adjustment of the proportional coefficient under abnormal operating conditions or short-term fluctuations, thereby ensuring the stability of the filter window length mapping relationship.

[0112] Completion ratio coefficient After the update, the updated scaling factor is substituted back into the calculation formula for the current filter window length to redetermine the filter window length at subsequent times.

[0113] By adopting the above scheme and introducing a closed-loop feedback fine-tuning mechanism based on the actual fluctuation of the final output temperature, the proportional coefficient can be corrected online using the actual standard deviation of the filtered output temperature, thereby compensating for the deviation between the theoretical model and the actual furnace tube operating conditions and improving the stability of system operation.

[0114] Optionally, the instantaneous temperature sequence is filtered according to the current filter window length to obtain the final output temperature, including: Based on the current filter window length, a moving average filter is applied to the instantaneous temperature sequence to obtain the candidate output temperature at the current moment; If the current filter window length is inconsistent with the filter window length of the previous time step, a weighted update is performed based on the output temperature and candidate output temperature of the previous time step to obtain the final output temperature of the current time step. If the current filter window length is the same as the filter window length at the previous time step, the candidate output temperature is taken as the final output temperature at the current time step.

[0115] Specifically, in determining the current filter window length Then, the instantaneous temperature sequence is dynamically filtered based on the current filter window length to obtain the final output temperature at the current moment. .

[0116] For example, based on the current filter window length The instantaneous temperature values ​​corresponding to the current moment and several consecutive sampling moments preceding it are selected from the instantaneous temperature sequence to form the current filtering window. Then, a moving average filter is applied to each instantaneous temperature value within this filtering window to obtain the candidate output temperature for the current moment. Its expression is:

[0117] in, Indicates the first The instantaneous temperature value at each sampling time. This indicates the length of the filter window at the current moment. This represents the candidate output temperature calculated using the current filter window length at the current moment.

[0118] If the current filter window length is the same as the filter window length at the previous moment, it means that the current filter intensity has not changed. In this case, the candidate output temperature can be directly used as the final output temperature at the current moment.

[0119] If the current filter window length is different from the filter window length at the previous moment, it indicates that the current filter intensity has been adjusted. To avoid a jump in output temperature due to a sudden change in filter window length, this embodiment uses the output temperature from the previous moment as a reference. and the candidate output temperature at the current moment Perform a weighted update to obtain the final output temperature at the current moment. Its expression is:

[0120] in, For the preset weighting coefficients, and .

[0121] In some embodiments, It can be set to a fixed value, such as 0.6, 0.7, or 0.8; in other embodiments, It can also be adaptively set according to the current change in the filter window length. That is, the greater the change in the filter window length, the higher the weight assigned to the output temperature at the previous moment, so as to further reduce the sudden change in output.

[0122] By adopting the above scheme, the instantaneous temperature sequence is subjected to moving average filtering based on the current filtering window length, which can perform targeted smoothing of the output temperature according to the current signal quality and process status. At the same time, when the filtering window length changes, a weighted update mechanism of the previous output temperature and the current candidate output temperature is further introduced, which can reduce the output jump caused by the switching of filtering intensity, thereby improving the output stability while maintaining the continuity of the output process.

[0123] Optionally, the adaptive filtering method for the semiconductor furnace tube temperature data mentioned above further includes: If the real-time signal-to-noise ratio remains below the preset lower limit for a preset duration, the current signal quality is determined to be non-compliant with measurement requirements, and an alarm signal is output.

[0124] Specifically, the preset signal-to-noise ratio (SNR) lower limit is used to characterize the lowest signal quality boundary that can meet the requirements of reliable measurement. When the real-time SNR... When the signal-to-noise ratio drops below the preset lower limit, it indicates that the current temperature measurement signal quality is significantly lower than the target requirement.

[0125] For example, it can be determined according to the following formula:

[0126] in, This is the preset lower limit for the signal-to-noise ratio.

[0127] Furthermore, to avoid false alarms caused by momentary disturbances or short-term anomalies, this embodiment does not immediately trigger an alarm when the real-time signal-to-noise ratio first falls below the preset lower limit of the signal-to-noise ratio. Instead, it continuously determines the low signal-to-noise ratio state, and only when the duration of the low signal-to-noise ratio state exceeds the preset duration is it determined that the current signal quality does not meet the measurement requirements and an alarm signal is output.

[0128] The aforementioned adaptive filtering method for semiconductor furnace tube temperature data calculates the instantaneous temperature sequence and real-time signal-to-noise ratio (SNR) based on the digital voltage signal, calculates the target SNR based on a preset noise threshold, and then dynamically determines the current filtering window length based on the real-time SNR, target SNR, and current process parameters. The instantaneous temperature sequence is then filtered based on the adjusted filtering window length. This application introduces a preset noise threshold set based on the process requirements of the current furnace tube process window. This quantifies the upper limit of the allowable deviation of the temperature measurement result fluctuation for engineers as the basis for filtering adjustment, ensuring that the filtering control process is simultaneously constrained by both the current signal quality and process requirements. Therefore, when the temperature measurement signal quality is poor, the filtering effect can be enhanced to reduce temperature output fluctuations, thereby improving measurement accuracy and stability. When the temperature measurement signal quality is good or the process has high requirements for dynamic tracking, the filtering effect can be weakened to reduce output hysteresis, thereby improving dynamic response performance. This achieves synergistic optimization of dynamic response performance while ensuring measurement accuracy and stability, and improves the adaptability of semiconductor furnace tube temperature measurement under different process windows.

[0129] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0130] Based on the same inventive concept, this application also provides an adaptive filtering system for semiconductor furnace tube temperature data. This system is applicable to the above-mentioned adaptive filtering method for semiconductor furnace tube temperature data. The solution provided by this system is similar to the solution described in the above method. Therefore, the specific limitations of one or more system embodiments provided below can be found in the limitations of the method above, and will not be repeated here.

[0131] Please see Figure 5 In one embodiment, the adaptive filtering system for semiconductor furnace tube temperature data includes: an acquisition module, a calculation module, and a filtering module.

[0132] The acquisition module is used to acquire digital voltage signals of two preset wavelength band measurement optical signals, preset noise thresholds, and current process parameters. The digital voltage signals are obtained by synchronously sampling two reflected lights on the wafer surface inside the furnace tube. The reflected lights are generated based on the optical signals of two preset wavelength bands irradiating the wafer surface. The preset noise thresholds are set based on the process requirements of the current furnace tube process window.

[0133] The calculation module is used to calculate the instantaneous temperature sequence based on the digital voltage signal, and calculate the real-time signal-to-noise ratio based on the instantaneous temperature sequence; calculate the target signal-to-noise ratio based on the preset noise threshold; and determine the current filtering window length based on the real-time signal-to-noise ratio, the target signal-to-noise ratio, and the current process parameters.

[0134] The filtering module is used to filter the instantaneous temperature sequence according to the current filtering window length to obtain the final output temperature.

[0135] Optionally, the calculation module calculates the instantaneous temperature sequence based on the digital voltage signal, including: calculating the corresponding two-color ratio based on two digital voltage signals at different sampling times; calculating the instantaneous temperature value corresponding to each two-color ratio based on the pre-stored calibration relationship between temperature and ratio; and sorting each instantaneous temperature value according to the sampling time order to obtain the instantaneous temperature sequence.

[0136] Optionally, the calculation module calculates the real-time signal-to-noise ratio based on the instantaneous temperature sequence, including: processing the instantaneous temperature sequence in real time using the sliding window method to obtain the window temperature set corresponding to the current moment; calculating the standard deviation of each instantaneous temperature value in the window temperature set relative to the window mean to obtain the temperature noise standard deviation; and calculating the real-time signal-to-noise ratio at the current moment based on the preset scaling factor and the temperature noise standard deviation, wherein the real-time signal-to-noise ratio is negatively correlated with the temperature noise standard deviation.

[0137] Optionally, the calculation module calculates the target signal-to-noise ratio based on a preset noise threshold, including: determining the target standard deviation based on a preset proportional relationship between the preset noise threshold and the noise standard deviation; and calculating the ratio of a preset scaling factor to the target standard deviation to obtain the target signal-to-noise ratio.

[0138] Optionally, the calculation module, when the preset window length threshold includes a preset minimum window length and a preset maximum window length, determines the current filtering window length based on the real-time signal-to-noise ratio, the target signal-to-noise ratio, and the current process parameters, including: The candidate filtering window length is determined by multiplying the process stage coefficient, the quartz window cumulative deposition factor, the proportionality coefficient, and the ratio of the target signal-to-noise ratio to the real-time signal-to-noise ratio. Based on the preset minimum and maximum window lengths, the candidate filtering window length is limited to obtain the current filtering window length, expressed as:

[0139] in, The preset maximum window length; Set the minimum window length; This is the proportionality coefficient; This is the process stage coefficient; For the quartz window, the accumulation deposition factor; The target signal-to-noise ratio; For real-time signal-to-noise ratio; It is a non-zero positive number, used to avoid division by zero; This indicates rounding down. Specifically, based on a preset minimum window length and a preset maximum window length, the candidate filter window length is limited to obtain the current filter window length. This includes: if the candidate filter window length is less than the preset minimum window length, the preset minimum window length is used as the current filter window length; if the candidate filter window length is greater than the preset maximum window length, the preset maximum window length is used as the current filter window length; and if the candidate filter window length is between the preset minimum and preset maximum window lengths, the candidate filter window length is used as the current filter window length.

[0140] Optionally, the calculation module is also used to acquire the data of the final output temperature within the most recent monitoring window according to a preset monitoring cycle, and calculate the output standard deviation based on the data; if the output standard deviation is greater than a first preset multiple of the target standard deviation, the scaling factor is increased according to a preset ratio; if the output standard deviation is less than a second preset multiple of the target standard deviation and the real-time signal-to-noise ratio is continuously lower than the target signal-to-noise ratio, the scaling factor is decreased according to a preset ratio; otherwise, the scaling factor remains unchanged; and the filter window length for subsequent times is re-determined based on the adjusted scaling factor.

[0141] Optionally, the filtering module filters the instantaneous temperature sequence according to the current filtering window length to obtain the final output temperature, including: performing a moving average filter on the instantaneous temperature sequence according to the current filtering window length to obtain the candidate output temperature at the current moment; if the current filtering window length is inconsistent with the filtering window length at the previous moment, performing a weighted update based on the output temperature at the previous moment and the candidate output temperature to obtain the final output temperature at the current moment; if the current filtering window length is consistent with the filtering window length at the previous moment, using the candidate output temperature as the final output temperature at the current moment.

[0142] Optionally, the adaptive filtering system for the semiconductor furnace tube temperature data also includes an alarm module.

[0143] The alarm module is used to determine that the current signal quality does not meet the measurement requirements and output an alarm signal if the real-time signal-to-noise ratio is continuously lower than the preset lower limit of the signal-to-noise ratio within a preset time period.

[0144] The aforementioned adaptive filtering system for semiconductor furnace tube temperature data calculates the instantaneous temperature sequence and real-time signal-to-noise ratio (SNR) based on the digital voltage signal, calculates the target SNR based on a preset noise threshold, and then dynamically determines the current filtering window length based on the real-time SNR, target SNR, and current process parameters. The instantaneous temperature sequence is then filtered based on the adjusted filtering window length. This application introduces a preset noise threshold set based on the process requirements of the current furnace tube process window. This quantifies the upper limit of the allowable deviation of the temperature measurement result fluctuation for engineers as the basis for filtering adjustment, ensuring that the filtering control process is simultaneously constrained by both the current signal quality and process requirements. Therefore, when the temperature measurement signal quality is poor, the filtering effect can be enhanced to reduce temperature output fluctuations, thereby improving measurement accuracy and stability. When the temperature measurement signal quality is good or the process has high requirements for dynamic tracking, the filtering effect can be weakened to reduce output hysteresis, thereby improving dynamic response performance. This achieves synergistic optimization of dynamic response performance while ensuring measurement accuracy and stability, and improves the adaptability of semiconductor furnace tube temperature measurement under different process windows. Each module in the aforementioned adaptive filtering system for semiconductor furnace tube temperature data can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of a computer device in hardware form or independent of it, or stored in the memory of the computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0145] In one feasible embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method steps in the adaptive filtering method for semiconductor furnace tube temperature data described above.

[0146] In one feasible embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the method steps in the adaptive filtering method for semiconductor furnace tube temperature data described above.

[0147] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0148] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0149] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. An adaptive filtering method for semiconductor furnace tube temperature data, characterized in that, The method includes: The system acquires a digital voltage signal, a preset noise threshold, and current process parameters. The digital voltage signal is obtained by photoelectric conversion and synchronous sampling of two reflected lights formed by two preset wavelength light signals irradiating the wafer surface inside the furnace tube. The preset noise threshold is set based on the process requirements of the current furnace tube process window. The current process parameters include process stage coefficients, quartz window cumulative deposition factor, and scaling factor. Calculate the instantaneous temperature sequence based on the digital voltage signal, and calculate the real-time signal-to-noise ratio based on the instantaneous temperature sequence; Calculate the target signal-to-noise ratio based on the preset noise threshold; The current filtering window length is determined based on the real-time signal-to-noise ratio, the target signal-to-noise ratio, and the current process parameters. The current filtering window length is associated with the ratio of the target signal-to-noise ratio to the real-time signal-to-noise ratio, and the current filtering window length is constrained and adjusted based on a preset window length threshold range. The instantaneous temperature sequence is filtered based on the adjusted current filter window length to obtain the final output temperature.

2. The method according to claim 1, characterized in that, The step of calculating the instantaneous temperature sequence based on the digital voltage signal includes: Calculate the corresponding two-color ratio based on the two digital voltage signals at different sampling times; Based on the pre-stored calibration relationship between temperature and ratio, calculate the instantaneous temperature value corresponding to each of the two-color ratios; The instantaneous temperature values ​​are sorted according to the sampling time order to obtain the instantaneous temperature sequence.

3. The method according to claim 1, characterized in that, The calculation of the real-time signal-to-noise ratio based on the instantaneous temperature sequence includes: The instantaneous temperature sequence is processed in real time using the sliding window method to obtain the set of window temperatures corresponding to the current moment; The standard deviation of each instantaneous temperature value in the window temperature set relative to the window mean is calculated to obtain the temperature noise standard deviation; The real-time signal-to-noise ratio is calculated based on the preset scaling factor and the temperature noise standard deviation. The real-time signal-to-noise ratio is negatively correlated with the temperature noise standard deviation.

4. The method according to claim 3, characterized in that, The step of calculating the target signal-to-noise ratio based on the preset noise threshold includes: The target standard deviation is determined based on the preset proportional relationship between the preset noise threshold and the noise standard deviation; The target signal-to-noise ratio is obtained by calculating the ratio of the preset scaling factor to the target standard deviation.

5. The method according to claim 1, characterized in that, The preset window length threshold includes a preset minimum window length and a preset maximum window length; The step of determining the current filter window length based on the real-time signal-to-noise ratio, the target signal-to-noise ratio, and the current process parameters includes: The candidate filter window length is determined by multiplying the process stage coefficient, the quartz window cumulative deposition factor, the proportional coefficient, and the ratio of the target signal-to-noise ratio to the real-time signal-to-noise ratio. Based on the preset minimum window length and the preset maximum window length, the candidate filter window length is subjected to amplitude limiting to obtain the current filter window length, which is expressed as: in, The preset maximum window length; Set the minimum window length; This is the proportionality coefficient; This is the process stage coefficient; For the quartz window, the accumulation deposition factor; The target signal-to-noise ratio; For real-time signal-to-noise ratio; It is a non-zero positive number, used to avoid division by zero; This indicates rounding down to the nearest integer.

6. The method according to claim 5, characterized in that, The step of limiting the candidate filter window length based on a preset minimum window length and a preset maximum window length to obtain the current filter window length includes: If the candidate filter window length is less than the preset minimum window length, the preset minimum window length shall be used as the current filter window length. If the candidate filter window length is greater than the preset maximum window length, the preset maximum window length shall be used as the current filter window length. If the candidate filter window length is between the preset minimum window length and the preset maximum window length, the candidate filter window length is taken as the current filter window length.

7. The method according to claim 4, characterized in that, The method further includes: According to the preset monitoring cycle, acquire the data of the final output temperature within the most recent monitoring window, and calculate the output standard deviation based on the data; If the output standard deviation is greater than a first preset multiple of the target standard deviation, the proportional coefficient is increased according to a preset ratio. If the output standard deviation is less than a second preset multiple of the target standard deviation and the real-time signal-to-noise ratio is continuously lower than the target signal-to-noise ratio, the proportional coefficient is reduced according to a preset ratio. Otherwise, keep the aforementioned proportionality coefficient unchanged; Based on the adjusted scaling factor, the filter window length for subsequent time steps is redetermined.

8. The method according to claim 1, characterized in that, The step of filtering the instantaneous temperature sequence according to the current filtering window length to obtain the final output temperature includes: Based on the current filter window length, the instantaneous temperature sequence is subjected to a moving average filter to obtain the candidate output temperature at the current moment; If the current filter window length is inconsistent with the filter window length of the previous time step, a weighted update is performed based on the output temperature of the previous time step and the candidate output temperature to obtain the final output temperature of the current time step. If the current filter window length is the same as the filter window length at the previous moment, the candidate output temperature is taken as the final output temperature at the current moment.

9. The method according to claim 1, characterized in that, The method further includes: If the real-time signal-to-noise ratio remains below the preset lower limit for a preset duration, it is determined that the current signal quality does not meet the measurement requirements, and an alarm signal is output.

10. An adaptive filtering system for semiconductor furnace tube temperature data, characterized in that, The system includes: The acquisition module is used to acquire digital voltage signals of two preset wavelength band measurement optical signals, a preset noise threshold, and current process parameters; wherein, the digital voltage signals are obtained by synchronously sampling two reflected lights on the wafer surface inside the furnace tube, and the reflected lights are generated based on the optical signals of two preset wavelength bands illuminating the wafer surface; the preset noise threshold is set based on the process requirements of the current furnace tube process window; The calculation module is used to calculate the instantaneous temperature sequence based on the digital voltage signal, and calculate the real-time signal-to-noise ratio based on the instantaneous temperature sequence; calculate the target signal-to-noise ratio based on the preset noise threshold; and determine the current filter window length based on the real-time signal-to-noise ratio, the target signal-to-noise ratio, and the current process parameters. The filtering module is used to filter the instantaneous temperature sequence according to the current filtering window length to obtain the final output temperature.

11. An in-situ measuring device, characterized in that, The device includes: The light source module is used to generate measurement light, separate the measurement light into two preset wavelength bands, and irradiate the wafer surface inside the furnace tube. The acquisition module is used to simultaneously sample two reflected lights generated on the wafer surface based on the optical signal and convert them into digital voltage signals; The adaptive filtering system for semiconductor furnace tube temperature data as described in claim 10 is configured to calculate an instantaneous temperature sequence based on the digital voltage signal, and calculate a real-time signal-to-noise ratio based on the instantaneous temperature sequence; calculate a target signal-to-noise ratio based on the preset noise threshold; determine a current filtering window length based on the real-time signal-to-noise ratio, the target signal-to-noise ratio, and the current process parameters; and filter the instantaneous temperature sequence based on the current filtering window length to obtain the final output temperature.

12. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1-9.

13. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1-9.

Images